Unable to train Multilabel Image Classifier in Sagemaker console. No option to set "AttributeNames".

**ERROR:**

_ClientError: Got empty batch from pipe. Please check if the manifest file contains valid batches and the value of training job parameter AttributeNames matches attributes names in the augmented manifest._

Using amazon SageMaker console Create training job option. I set up the parameters with the help of [https://docs.aws.amazon.com/sagemaker/latest/dg/image-classification.html](image-classification docs).

> Channel name: train, validation (both with the same options, seperate .manifest files)
> Input mode: Pipe
> Content type: application/x-image
> Record Wrapper: RecordIO
> Data source: S3 bucket

I used an augmented manifest file created by amazon sagemaker Ground Truth.

According to the error I need to set AttributeNames to the Ground Truth job name, yet there is no such option to do so. AND amazon keeps billing me for p2 instance time yet again. I feel like I'm being robbed using sagemaker.

The report names no repository file, notebook, or test; start by reviewing the SageMaker console Create training job flow and the linked image-classification documentation. Compare the augmented manifest requirements and the AttributeNames training parameter, then determine whether the missing console option belongs in this examples repository or requires an AWS service report.

Written by the indexing model from the issue text.